Assignment 8

a. Compute Combat Power (15 pts)

Again, the maximum combat power can be computed via information here for level 40 as:

Attack = 2 * round(attack^0.5 * sp_attack^0.5 + speed^0.5)
Defense = 2 * round(defense^0.5 * sp_defense^0.5 + speed^0.5)
Stamina = 2 * hp
MaxCP = (Attack + 15) * (Defense + 15)^0.5 * (Stamina + 15)^0.5 * 0.7903001^2 / 10

We wish to create a dataframe with a new column named max_cp that stores this computed value. Do this without loops. Remember that we can do computations with series (columns) in dataframes.

b. Maximum Combat Power by Generation (10 pts)

Now, find the Pokémon with the greatest maximum combat power in each of the eight generations. Output the name of the pokemon along with it’s max_cp value.

a Mean Values by Primary Type (10 pts)

Create a dataframe that groups by the primary type and computes the mean attack, defense, and speed for each primary type.

b. Melt/Unpivot the Dataframe (5 pts)

Next, melt/unpivot this data to put it into a long format with three columns: primary_type, variable, and value, where variable is one of attack, defense, or speed, and value is the corresponding mean value.

c. Bar Chart (10 pts)

Using seaborn, matplotlib, or via pandas’ plotting routines, create a grouped bar chart that shows the mean attack, defense, and speed for each primary type. Set the figure size wider to improve the visualization. Include a legend and make sure the axes are properly labeled.

With either pandas or polars, seaborn’s barplot is the most straightforward solution. Use the penguins grouped by sex as a guide.

If you are using pandas, you can use its built-in plot method (which uses matplotlib) to create the grouped bar chart, but you will want to go back to the unmelted version of the dataset. Make sure the rows are primary types and the columns are the three statistics.

If you use polars, you can use to_pandas to convert to pandas and use its plot facilities as described above, but you can do this directly via matplotlib, although this is much more involved. You will receive extra credit for doing it this way. The easiest way to draw a grouped bar chart is to draw three different bar charts, one for each statistic, but with proper offsets for the x positions. Your loop body should be a call to pyplot. The offset calculation is a bit tricky so here is a starting structure for how this works:

import matplotlib.pyplot as plt

# convert the category to a number for offsets
df = df.with_columns(
    pl.col("primary_type").cast(pl.Categorical).cast(pl.UInt32).alias("primary_num")
)
for i, col in enumerate(["attack", "defense", "speed"]):
    df_offset = df.with_columns((pl.col("primary_num") * 2 + (i - 1) / 2).alias("offset"))
    plt.bar(
        ...,
        data=df_offset
    )
# other calls for legend, labels

d. Scatterplot (15 pts)

From the bar chart, we think there may be a negative correlation between attack and defense averages versus speed averages. Using the grouped data from Part a, and then add the attack and defense means together to create a new column. Using matplotlib directly (seaborn and pandas are not allowed), plot this new column versus the speed column. Label the axes appropriately. You should see one significant outlier with respect to the negative correlation. Using values from the visualization, write a filter operation to identify the primary type of this outlier. (You should be able to estimate the values from the plot axes to construct a filter.)

e. [CS503 Only] Scatter Matrix (15 pts)

Now, let’s use all three attributes and use altair to create a scatter matrix. A scatter matrix is a bunch of scatterplots, one for each pair of variables. Begin by creating a scatterplot that compares just attack and speed. Once you have this, change the x and y values to use the repeat capabilities of altair, and set the repeat over the speed, attack, and defense attributes.

a. Bubble Chart (15 pts)

Using altair, create a bubble chart visualization that shows Pokémon positioned by attack and defense, sized by speed, and colored by generation. Make sure you aren’t biasing the visualization toward showing more from the later generations; this occurs by having later generations plotted on top of the older generations.

b. Binned Scatterplot (15 pts)

Since Part b led to major overdraw problems (points occluding other points), we will create a binned scatterplot that shows the count of the number of Pokémon in a particular attack/defense bin. This will help us better see the distribution of Pokémon according to attack and defense measures. Use 1200 bins (40x30) or some other reasonable number that shows the trend.